Dave Chinner [Fri, 15 Mar 2013 11:53:25 +0000 (11:53 +0000)]
xfstest: enable xfs_repair for large filesystem testing
Now that large filesystem testing does not play free space games to
fill the space without IO, we can enable xfs_repair when running in
this mode. xfs_repair has had it's scalability problems solved, too,
so this is a safe thing to do.
Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Rich Johnston <rjohnston@sgi.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Dave Chinner [Fri, 15 Mar 2013 11:53:24 +0000 (11:53 +0000)]
xfstests: use command line option for setting extra space
Allow the extra free space to leave in large scratch filesystems to
be specified by a command line option rather than just via an
environment variable.
Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Rich Johnston <rjohnston@sgi.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Dave Chinner [Fri, 15 Mar 2013 11:53:23 +0000 (11:53 +0000)]
xfstests: use preallocation for ag-wiper
To enable sane testing of large scale filesystems, the --large-fs
test option uses xfs_db magic to mark AGs full without doing any IO.
This leaves only a small amount of free space left in the filesystem
to stress the high AGs of the filesystem rather than the low AGs.
This method requires us to have special filesystem check options to
avoid free space checking in xfs_check, and we cannot current run
xfs_repair on such a filesystem at all. As it is, free space
checking on xfs_check does not scale, so we still need to avoid this
checking regardless of how we fill the filesystem.
We can achieve exactly the same fill behavior by preallocating a
single large file in the filesystem immediately after creating it.
This is a filesystem independent manner of filling the filesystem,
and allows us to do large filesystem testing on more than just XFS.
Further, this preallocation method effectively adds a new "very
large file" test. It also enables us to run an unmodified xfs_repair
or filesystem specific fsck program to check the filesystem for
sanity, so we can now do full sanity checking of such large
filesystems.
Dave Chinner [Fri, 15 Mar 2013 11:53:22 +0000 (11:53 +0000)]
xfstests: rename RETAIN_AG_BYTES
Rename the $RETAIN_AG_BYTES variable to be more generic so that it
reflects the fact that it is designed to retain a certain amount of
extra free space above the default amount in the filesystem when
doing large scratch device testing.
Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Rich Johnston <rjohnston@sgi.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Dave Chinner [Fri, 15 Mar 2013 11:53:21 +0000 (11:53 +0000)]
xfstests: rename USE_BIG_LOOPFS to be more generic
USE_BIG_LOOPFS is really misnamed - it can be used on real devices just as
easily as loop devices. It really means we are testing a large scratch device
and that we should enable the special filesystem filling and checking options
that enable xfstests to be run sanely on large XFS filesystems.
Dave Chinner [Fri, 15 Mar 2013 11:53:20 +0000 (11:53 +0000)]
xfstests: add --largefs check option
Make it easier to check large filesystems quickly by adding a
--large-fs option to check to turn on shortcuts for large scratch
device filesystem testing.
Also, reject invalid command line options with a usage message.
Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Rich Johnston <rjohnston@sgi.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Eric Sandeen [Wed, 13 Mar 2013 19:01:58 +0000 (19:01 +0000)]
xfstests: btrfs tests for basic informational commands
Run basic btrfs information commands in various ways, performing
sanity tests of: filesystem show, label, sync, and device stats
(sync is included just because it's simple). These are mostly
just smoke tests, although for example show by label & UUID
should verify that the correct fs was shown.
This also adds quite a few new filters to accommodate the output
of the new commands.
Cc: linux-btrfs@vger.kernel.org Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Eric Sandeen [Wed, 13 Mar 2013 19:01:58 +0000 (19:01 +0000)]
xfstests: btrfs tests for basic informational commands
Run basic btrfs information commands in various ways, performing
sanity tests of: filesystem show, label, sync, and device stats
(sync is included just because it's simple). These are mostly
just smoke tests, although for example show by label & UUID
should verify that the correct fs was shown.
This also adds quite a few new filters to accommodate the output
of the new commands.
Cc: linux-btrfs@vger.kernel.org Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Stefan Behrens <sbehrens@giantdisaster.de> Signed-off-by: <rjohnston@sgi.com>
If we receive an unexpected result from an async write, the error
reporting does not tell the actual number of bytes written. Fix that,
and also a couple of typos in printf's.
Eric Sandeen [Tue, 5 Mar 2013 22:41:54 +0000 (22:41 +0000)]
xfstests: Fix hang when mkfs.btrfs isn't present
My earlier patch (xfstests: handle new mkfs.btrfs -f option cleanly)
had a flaw in that if set_prog_path mkfs.btrfs returns nothing,
the grep will hang.
Test for that case to avoid it, and just return the empty string
in that case.
David Sterba [Wed, 23 Jan 2013 11:14:33 +0000 (11:14 +0000)]
xfstests: make length of diff output configurable
In commit 11c1d79414e2571 "xfstests: Change the diff output of failed
tests", the diff output of a failed test was hardcoded to 10 lines to
avoid overly long output and user can get the full output by manually
running the diff. However this is not always possible and convenient,
eg. in repeated automated tests where the required information is lost
after the test round finished. Then the caputred logs do not contain
enough informatin for analysis.
Introduce the DIFF_LENGTH env variable to tune the diff size, keeping it
10 as deafult and 0 to disable the limit.
Signed-off-by: David Sterba <dsterba@suse.cz> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Eric Sandeen [Tue, 26 Feb 2013 20:42:38 +0000 (20:42 +0000)]
xfstests: handle new mkfs.btrfs -f option cleanly
I added an "-f" option to mkfs.btrfs to force overwrite
of an existing filesystem. Now on an xfstests run, new
mkfs.btrfs requires it, and old mkfs.btrfs cannot accept
it.
So, add a helper which works out whether -f is needed,
and add it to the MKFS_BTRFS_PROG env. var as necessary,
so that it is an always-included option during the tests.
Eric Whitney [Tue, 26 Feb 2013 21:03:22 +0000 (21:03 +0000)]
xfstests: use native definition of O_DIRECT flag
The definition of O_DIRECT in src/trunc.c causes xfstest 125 to fail
when run on a Pandaboard. On ARM, the value used (0x040000) is
O_DIRECTORY rather than O_DIRECT as it is on x86. Prefer the platform's
native definition of O_DIRECT supplied by fcntl.h if available. Also,
fix a couple of error messages to properly reflect their context.
Signed-off-by: Eric Whitney <enwlinux@gmail.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Eric Whitney [Tue, 5 Mar 2013 19:17:48 +0000 (19:17 +0000)]
xfstests: don't remove trailing zeros from integers
_within_tolerance strips trailing zeros from the min and max range
values it outputs. This leads to damage if the min or max value is
an integer containing trailing zeros rather than a real number with
a fractional part containing trailing zeros. Xfstest 289 can exhibit
this problem when its input is out of range. Modify the code so it
will only remove trailing zeros found after a decimal point.
V1->V2: Remove decimal points not followed by digits
V2->V3: Per Dave Chinner, simplify by using multiple sed expressions
Signed-off-by: Eric Whitney <enwlinux@gmail.com> Reviewed-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Theodore Ts'o [Tue, 5 Mar 2013 17:59:42 +0000 (17:59 +0000)]
xfstests: don't assume that falloc_punch implies falloc in test 255
As of Linux 3.9-rc1, ext4 will support the punch operation on file
systems using indirect blocks, but it can not support the fallocate
operation (since there is no way to mark a block as uninitialized
using indirect block scheme). This caused test 255 to fail, since it
only used _require_xfS_io_falloc_punch assuming that all file systems
which supported punch can also support fallocate. Fix this.
Commit 05dadc0 ("Btrfs: add fiemap's flag check") added validity
checks to the fiemap flags and hence invalid flags are now being
rejected. Test 276 passes an invalid fiemap flag to btrfs, and so
that test fails since this commit was added.
Btrfs doesn't support FIEMAP_FLAG_XATTR, which is enabled by -x option
of filefrag, and will fail with
'FIBMAP: Invalid argument'
for 'filefrag -vx'. 'filefrag -vx' fails on btrfs with
'FIEMAP failed with unsupported flags 2'
Remove the '-x' option.
Signed-off-by: Wang Sheng-Hui <shhuiw@gmail.com> Reviewed-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:15 +0000 (10:42 +0000)]
xfstests: add disk failure simulation test
There are many situations where disk may fail for example
1) brutal usb dongle unplug
2) iscsi (or any other netbdev) failure due to network issues
In this situation filesystem which use this blockdevice is
expected to fail(force RO remount, abort, etc) but whole system
should still be operational. In other words:
1) Kernel should not panic
2) Memory should not leak
3) Data integrity operations (sync,fsync,fdatasync, directio) should fail
for affected filesystem
4) It should be possible to umount broken filesystem
Later when disk becomes available again we expect(only for journaled filesystems):
5) It will be possible to mount filesystem w/o explicit fsck (in order to catch
issues like https://patchwork.kernel.org/patch/1983981/)
6) Filesystem should be operational
7) After mount/umount has being done all errors should be fixed so fsck should
not spot any issues.
This test use fault injection (CONFIG_FAULT_INJECTION=y,
CONFIG_FAIL_MAKE_REQUEST=y and CONFIG_FAULT_INJECTION_DEBUG_FS=y config
options) to force all new IO requests to fail for a given device. Xfs
already has XFS_IOC_GOINGDOWN ioctl which provides similar behavior, but it
is fs specific and it does it in an easy way because it performs freeze_bdev()
before actual shutdown.
Test run fsstress in background and then force disk failure.
Once disk failed it check that (1)-(4) is true.
Then makes disk available again and check that (5)-(7) is also true
BE CAREFUL!! test known to cause memory corruption for XFS
see: https://gist.github.com/dmonakhov/4953045
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:14 +0000 (10:42 +0000)]
xfstest: add defragmentation stress tests for ext4
Perform various regression tests for ext4defrag subsystem
301'th Test1: Defragment file while other task does direct AIO
302'th Test2: Perform defragmentation on file under buffered AIO
while third task does direct AIO to donor file
303'th Test3: Two defrag tasks use common donor file.
304'th Test4: Stress defragmentation. Several threads perform
fragmentation at random position use inplace=1 will
allocate and free blocks inside defrag event improve
load pressure.
This tests are considered dangerous because 302'th and 303'th are known
to trigger OOPS on recent kernels see:https://gist.github.com/dmonakhov/4770294
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:13 +0000 (10:42 +0000)]
xfstest: add fallocate/punch_hole vs AIO/DIO stress test
Run random AIO/DIO activity (fio's job:direct_aio_raicer)
random fallocate activity(fio's job:falloc_raicer)
and random punch_hole activity(punch_hole_raicer) on a common
file in parallel. If a race exists, old dio request may rewrite
punched block after it was allocated to another file, we will
catch that by verifier fio's job: "aio-dio-verifier".
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:12 +0000 (10:42 +0000)]
xfstest: add fallocate/truncate vs AIO/DIO stress test
Run DIO, fallocate and truncate threads on a common file in parallel.
If a race exists, the old dio request may rewrite blocks after it was
allocated to another file, we will catch that by verifying blocks content.
this patch known to catch deadlock for ext4
http://lists.openwall.net/linux-ext4/2012/09/06/3
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:10 +0000 (10:42 +0000)]
xfstest: allow fsstress to use load factor where appropriate
1) Add _scale_fsstress_args function which transform arguments according
to load factors
2) Let all non deterministic fsstress tests to use scaled arguments
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:08 +0000 (10:42 +0000)]
xfstests: hardcode fops for determinable fsstests runs
106,107 and 117 are frozen tests which use known seed, it is
reasonable to explicitly hardcode file operations in order to avoid
implicit changes caused by future changes in fsstress.
NOTE: options generated like follows: fsstress -S c $ORIG_ARGS
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:07 +0000 (10:42 +0000)]
xfstest: add configurable load factors
Most stress test has probable behavior, the longer test run the
larger corner cases will be cover. It is reasonable to allow
user to provide some sort of system load factor.
This patch introduce two global variables
LOAD_FACTOR: Usually means factor number of running tasks
TIME_FACTOR: Usually means factor of run time, or number of operations
If not specified both variables defined to 1, so original behavior
preserved.
TODO: Change all stress tests to use this variables
Dmitry Monakhov [Wed, 20 Feb 2013 10:42:06 +0000 (10:42 +0000)]
xfstests: add fio requirement V2
FIO is very flexible io generator, I would call it IO swiss knife.
Currently we have tons of hardcoded application which reproduces
some predefined scenario. This approach has obvious disadvantages
1) Lack of flexibility: one written it is hard to modify it in future
2) Code base is large, many routines written again and again
At the same time add new fio based test, just add simple INI file.
This greatly simplifies code review. I do believe that some day we will
replace most of hardcoded io binaries with fio.
One who is planning to run $FIO_PROG should first check that system
contains appropriate version which is able to handle jobfile
for example: _require_fio 286-job.fio
Tomas Racek [Sun, 3 Feb 2013 10:19:58 +0000 (10:19 +0000)]
xfstests: 298: Test that FS sends discard requests only on free blocks
V1->V2: Change way of testing suggested by Dave Chinner
1. Create image file with FS on it
2. Call fstrim to discard free blocks
3. Check that every punched hole in the image file is in the area
that is marked as free
Andrew Dahl [Mon, 14 Jan 2013 18:17:09 +0000 (12:17 -0600)]
xfstests: Refactor release scripts to conform to using git archive
Refactored release scripts to conform to using git archive
When generating a release, there is a risk of some files being
stale, such as configure and the m4/autotools temp files. This
is fixed with a clean at the beginning of release generation.
In addition, there is no uniformity in the current method of
source tar generation between xfs utilities. Using git archive
solves this issue across all utilities.
Signed-off-by: Andrew Dahl <adahl@sgi.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Reported-by: fugazzi® <fugazzi99@gmail.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Eric Sandeen [Wed, 2 Jan 2013 22:28:32 +0000 (16:28 -0600)]
xfstests: test multiply-logged inodes & continued transactions
New test for 2 xfs_logprint error cases:
As xlog_print_trans_inode() stands today, it will error
out if more than one flag is set on f->ilf_fields:
xlog_print_trans_inode: illegal inode type
but this is a perfectly valid case, to have i.e. a data and
an attr flag set.
and:
xlog_print_trans_inode() has a special case for 2
specific op_head->oh_len lengths. If it matches
sizeof(xfs_inode_log_format_32_t) or
sizeof(xfs_inode_log_format_64_t), it assumes that
it's got an inode, and attempts to convert it and
print it accordingly.
However, if we arrive here via an op header which
is continued, then the length is simply a continuation
of the previous op, and it might *randomly* match the
size of one of the inode log formats, and thus get parsed
incorrectly.
Test both of these cases.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Satoru Takeuchi [Sun, 9 Dec 2012 23:37:41 +0000 (08:37 +0900)]
xfstests: Change the interpreter of ltp/runtest.sh to bash
In some environment, Running 080 fails with many errors.
It happens because ltp/rwtest.sh assumes /bin/sh is `bash' and it's not
always true (ex. /bin/sh is `dash' in Debian.) This script should
explicitly declare that its interpreter is `bash'.
Signed-off-by: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Eric Sandeen [Wed, 9 Jan 2013 16:17:59 +0000 (10:17 -0600)]
xfstests: test EROFS vs. EEXIST when creating on an RO filesystem
TBH, I don't know if this is posix-specified, but I found out the
hard way that when trying to re-create existing files on a readonly
filesystem, some apps expect/handle EEXIST, but fail on EROFS.
This will test mkdir, mknod, and symlinks for that behavior.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Eric Sandeen [Wed, 9 Jan 2013 16:13:26 +0000 (10:13 -0600)]
xfstests: move _filter_ln to common.filter
Move test 103's _filter_ln to common.filter and make
it more generic (not depending on 103's pathnames).
TBH I've lost my children's treasury of ln failure
messages, so I'm not sure this catches all variants;
it's hard to work backwards from the existing sed script
to what the various outputs were. This works for me
but might need more tweaking on other systems.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Currently the entire diff is printed to the stdout in the case that the
test output template differs from the actual test run. However in some
cases the diff can be _very_ long. This commit changes it so that we
print only first 10 lines of the diff.
Also indent the diff output from the left by four spaces for better
output readability.
Signed-off-by: Lukas Czerner <lczerner@redhat.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Jan Kara [Tue, 18 Dec 2012 10:35:50 +0000 (11:35 +0100)]
xfstests: Use qa_user and qa_group for test 219
Test 219 requires a special user. Use $qa_user and $qa_group (added in this
patch) for that purpose instead of hardcoded uid & gid. This also fixes
a false failure when repquota does not report quota for users not in passwd.
Signed-off-by: Jan Kara <jack@suse.cz> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Rich Johnston [Fri, 9 Nov 2012 13:24:01 +0000 (13:24 +0000)]
xfstests: resolve compiler warnings
This patch resolves the following compiler/configure warnings:
configure: WARNING: unrecognized options: --enable-lib64
warning: 'xxxx' may be used uninitialized in this function [-Wuninitialized]
warning: variable 'xxx' set but not used [-Wunused-but-set-variable]
warning: call to function 'foo' without a real prototype [-Wunprototyped-calls]
Signed-off-by: Rich Johnston <rjohnston@sgi.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
Rich Johnston [Fri, 9 Nov 2012 13:24:00 +0000 (13:24 +0000)]
xfstests: remove unused conditional NO_XFS
This patch removes the unused conditional NO_XFS which was added with
commit 332ee29d1ab7a8690ca8f6d06b0d116e3a4a4f0f:
"xfstests updates - rework build to be like other xfs packages ..."
Signed-off-by: Rich Johnston <rjohnston@sgi.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
sat [Mon, 19 Nov 2012 03:26:54 +0000 (03:26 +0000)]
xfstests: fix to build src/realpath and the correct the existence of target devices
Hi Rich, Eric
(2012/10/27 1:46), Rich Johnston wrote:
> On 06/08/2010 03:03 PM, Eric Sandeen wrote:
>> If you try running xfstests on lvm volumes which are symlinks,
>> it'll fail to run several tests because our _require_scratch
>> framework ultimately uses lstat not stat, and does not think
>> the lvm device (which is usually a symlink to a dm-X device)
>> is a block device. Sigh.
>>
>> Last try at this - just resolve any symlinked devicenames
>> into their realpath(3) in common.config.
>>
>> This actually seems to work.
>>
>> Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
>>
>> ---
>>
>>
>> diff --git a/common.config b/common.config
>> index 926846b..e5b2483 100644
>> --- a/common.config
>> +++ b/common.config
>
> Looks good
>
> Reviewed-by: Rich Johnston <rjohnston@sgi.com>
>
> Eric,
>
> This patch has been committed to git://oss.sgi.com/xfs/cmds/xfstests, master branch, commit ID d5ea873f.
Andrew Dahl [Wed, 14 Nov 2012 18:57:35 +0000 (18:57 +0000)]
xfstests: xfs_tosspages() test addition
Tests the XFS_IOC_ZERO_RANGE ioctl by way of the "xfs_io zero" utility
to ensure it is tossing the expected ranges.
The ranges tested are [0,1] [0,4095] [0,4096] [0,4097] [4095,8191]
[4095,8192] [4095,8193] [4096,8192] [1024,4096]
Signed-off-by: Andrew Dahl <adahl@sgi.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Mark Tinguely <tinguely@sgi.com>
wenqing.lz [Sun, 21 Oct 2012 13:15:09 +0000 (13:15 +0000)]
xfstests: get return value in seek_sanity_test
In seek_sanity_test getting return value is missing. Thus we couldn't
get the result unless someone read the log file (e.g. # 285).
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Reviewed-by: Jie Liu <jeff.liu@oracle.com> Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Rich Johnston [Tue, 6 Nov 2012 20:23:59 +0000 (14:23 -0600)]
Revert "xfstests: optionally run all tests under quota"
Signed-off-by: Rich Johnston <rjohnston@sgi.com> Reviewed-by: Dave Chinner <dchinner@redhat.com>
This reverts commit edd6bcf9c478036df9056df05924f791a0d322dd which caused
the following tests to fail:
142.out.bad:Can't initialize the DMAPI
143.out.bad:Can't initialize the DMAPI
144.out.bad:Can't initialize the DMAPI
145.out.bad:Can't initialize the DMAPI
146.out.bad:Can't initialize the DMAPI
147.out.bad:Can't initialize the DMAPI
150.out.bad:Can't initialize the DMAPI
155.out.bad:Can't initialize the DMAPI
156.out.bad:Can't initialize the DMAPI
157.out.bad:Can't initialize the DMAPI
158.out.bad:Can't initialize the DMAPI
161.out.bad:Can't initialize the DMAPI
168.out.bad:Can't initialize the DMAPI
175.out.bad:Can't initialize the DMAPI
176.out.bad:Can't initialize the DMAPI
185.out.bad:Can't initialize the DMAPI
Eric Sandeen [Thu, 25 Oct 2012 17:19:24 +0000 (17:19 +0000)]
xfstests: test ext4 statfs
Calculating free blocks in ext[234] is surprisingly hard, since
by default we report "bsd" style df which doesn't count filesystem
"overhead" blocks as used.
With a lot of code dedicated to sorting out what to report as
free, things tend to go wrong surprisingly often.
Here's a test to actually try to stop the next regression. ;)
NB: For bsddf, the kernel currently does not count journal blocks
as overhead; it probably should. But the test below looks to have
the result within 1% of perfection, so it still passes even if
the kernel doesn't count the journal against free blocks.
Eric Sandeen [Tue, 9 Feb 2010 19:26:31 +0000 (19:26 +0000)]
xfstests: optionally run all tests under quota
This patch might be a little heavy handed, but it seems to
work; if you set USE_QUOTA=1 in your environment, all
tests should be run with quota on and enabled.
This will hopefully help shake out some of the ext4
quota problems, although one needs to keep an eye
on the console to see whether warnings scroll by.
Eric Sandeen [Tue, 8 Jun 2010 20:03:39 +0000 (20:03 +0000)]
xfstests: resolve symlinked devices to real paths
If you try running xfstests on lvm volumes which are symlinks,
it'll fail to run several tests because our _require_scratch
framework ultimately uses lstat not stat, and does not think
the lvm device (which is usually a symlink to a dm-X device)
is a block device. Sigh.
Last try at this - just resolve any symlinked devicenames
into their realpath(3) in common.config.
Eric Sandeen [Thu, 16 Sep 2010 18:07:47 +0000 (18:07 +0000)]
xfstests: add local.config example
I think it'd be useful to have a local.config template in the
git repo, to avoid having to re-read the README each time
to remember the required environment variable names.
Thoughts? Am I just lazy? :)
Signed-off-by: Eric Sandeen <sandeen@sandeen.net> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Rich Johnston <rjohnston@sgi.com> Signed-off-by: Rich Johnston <rjohnston@sgi.com>
xfstests: Assume yes when test device is not partitioned
If an entire device used for file system creation, the mkfs.ext2/3/4
asks for confirmation before proceeding. Since the device is
configured for testing it is safe to assume positive response
during mkfs.
The patch also replaces hard coded mkfs path with MKFS_PROG variable.
Mike Frysinger [Mon, 24 Sep 2012 23:39:38 +0000 (19:39 -0400)]
xfstests: install shared libs with +x bits
These are shared libs w/executable code, so make sure they have +x bits
set on them. Some kernels will proactively disallow executable mmaps if
the files lack +x bits. It's also the right thing to do.
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Lukas Czerner [Tue, 16 Oct 2012 09:41:33 +0000 (09:41 +0000)]
xfstests: Test FITRIM where length is smaller than FSB
This tests corner case in FITRIM implementation where range size is
smaller than file system block or zero. In this case FITRIM should fail
with EINVAL.
The problem was spotted in xfs and ext4 where in case of length = 0 the
'end' variable underflowed. In case of length smaller than 1 FSB FITRIM
finished successfully, but we really should rather return EINVAL in both
cases.
(This patch has to be applied after 'Use upstream version of fstrim
instead of the local one')
Tomas Racek [Tue, 9 Oct 2012 13:24:10 +0000 (13:24 +0000)]
Use upstream version of fstrim instead of the local one
Local version of fstrim was dropped so that we depend on upstream
version which is now available via FSTRIM_PROG. _require_fstrim was
added to check if fstrim is available in the system and
_test_batched_discard to check if we can run fstrim on certain
mountpoint.
Also tests 251 and 260 were modified to reflect this change.
Eric Sandeen [Fri, 28 Sep 2012 03:33:45 +0000 (03:33 +0000)]
xfstests 068: Add mmap load
Test 068 is the go-to test for freeze deadlock coverage;
unfortunately it only uses fsstress, which doesn't do any
mmap IO.
Using the existing fstest binary gets us a cheap mmap
exerciser as well.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Jeff Liu [Sat, 2 Jun 2012 14:33:06 +0000 (14:33 +0000)]
xfstests: fix a compile warning at seek_sanity_test.c
Fix the compiler report warning at seek_sanity_test.c:
seek_sanity_test.c:46:3: warning: format '%lx' expects type 'long unsigned int', but argument 3 has type 'int'
Signed-off-by: Jie Liu <jeff.liu@oracle.com> Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Eric Sandeen [Wed, 19 Sep 2012 22:53:56 +0000 (17:53 -0500)]
xfstests: add _require_freeze and minor cleanups
More filesystems have grown freeze capability, so rather than
hardcoding several in _supported_fs, make tests 068 and 280
generic and then add a new _require_freeze() which checks whether
the fs under test can be frozen before beginning the test.
Minor other cleanups to 280:
- remove extra _supported_fs line
- clear $seq.full before beginning
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Tao Ma [Fri, 14 Sep 2012 11:03:12 +0000 (19:03 +0800)]
xfstests: 275, Don't remove all the files in SCRATCH_MNT.
After the volume is mkfsed, it should have no other files that
need to be rmed. And what's more, it makes ext4 complain since
lost+found is also removed.
So remove this useless "rm -rf $SCRATCH_MNT/*".
Signed-off-by: Tao Ma <boyu.mt@taobao.com> Reviewed-by: Eric Sandeen <sandeen@sandeen.net> Signed-off-by: Ben Myers <bpm@sgi.com>
Jan Kara [Thu, 16 Aug 2012 09:14:35 +0000 (11:14 +0200)]
xfstests: Make test 272 work for ext3
ext3 does not support direct IO for files with data journalling. This
confuses test 272. Make the test check whether open succeeds and perform
the writing only if it does.
Thanks for Dave Chinner for suggesting a simpler way to fix the test.
Signed-off-by: Jan Kara <jack@suse.cz> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Ben Myers <bpm@sgi.com>
Eric Sandeen [Wed, 12 Sep 2012 18:53:27 +0000 (13:53 -0500)]
285: Test xfs projid32bit functionality a bit more extensively
The test covers several areas including enabling projid32bit
functionality dynamically by xfs_admin, dumping, restoring, quota
reporting and xfs_db projid values reporting.
This test case hits a bug with xfsdump/xfsrestore process on a
projid32bit enabled filesystem.
Eric Sandeen: change {16,32}less filenames to {16,32}bit, add quick group
Signed-off-by: Boris Ranto <ranto.boris@gmail.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Eric Sandeen [Tue, 11 Sep 2012 01:46:10 +0000 (20:46 -0500)]
275: Fix to test and pass properly
This is a significant rework of 275, which made too many
assumptions about details of space usage and failed on several
filesystems (it passed on xfs, but only by accident).
This new version tries to leave about 256k free, then tries
a single 1M IO, and fails only if 0 bytes are written.
It also sends a lot more to $seq.full for debugging on failure,
doesn't rm -rf $SCRATC_MNT, and fixes a few other stylistic things.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com>
Rich Johnston [Tue, 28 Aug 2012 17:26:08 +0000 (17:26 +0000)]
xfstests: Use installed libraries only
The xfstests autotools currently searches locally in
../{acl,attr,dmapi,xfsprogs}
for libraries that xfstests depends upon, in addition to searching for them in
their regular installed locations on a system, e.g. /usr/lib. It appears this
feature was added (but not documented) so that xfs developers can build and run
xfstests without having to install the libraries. This can lead to trouble if
you expect that xfstests is using the versions of the libraries installed on
the system.
If a local library was found and not installed, libtool will create a wrapper
script to call the binary from the .libs directory. This patch will remove
searching for local libraries so that the installed libraries are always used.
Signed-off-by: Rich Johnston <rjohnston@sgi.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Ben Myer <bpm@sgi.com>
Rich Johnston [Tue, 28 Aug 2012 17:26:07 +0000 (17:26 +0000)]
xfstests: Use libtool to install applications
This patch replaces the INSTALL macro with LTINSTALL so that libtool is used
to install the applications. Libtool will install the binary from the proper
location and display a warning if any shared library dependancies are not
properly installed. This ensures that a libtool wrapper shell script is not
installed in place of application when libtool wrappers are being used.
Signed-off-by: Rich Johnston <rjohnston@sgi.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Ben Myer <bpm@sgi.com>
Dave Chinner [Thu, 26 Jul 2012 08:35:06 +0000 (08:35 +0000)]
xfstests: speed up 227 by using preallocation
Test 227 (fsr test) creates fragmented fre space by doing lots of
small writes to sparse offsets ni a file. This seeks the disk heads
around a lot writing data. We don't need to write data - just
trigger allocation. Hence use preallocation instead of data writes
and run at allocation speed rather than data write speed.
Dave Chinner [Thu, 26 Jul 2012 08:35:05 +0000 (08:35 +0000)]
xfstests: _check_quota_usage needs to unmount to get XFS quotacheck
Remount won't run a quota check - it's only done during mount. Hence
all quota tests using this check function are not actually
validating XFS filesystems right now.
Dave Chinner [Thu, 26 Jul 2012 08:35:04 +0000 (08:35 +0000)]
xfstests: loop devices vs umount stupidity
Unmounting a fileystem mounted on a loop device doesn't always tear
down the loop device. Its racy, and it causes tests to randomly
fail.
To avoid that, we have to use umount -d to ensure that we destroy
loop devices under filesystems in case the kernel doesn't tear it
down automatically to prevent the test from failing. However, if
the kernel does tear it down automatically, umount now issues a
warning that it couldn't tear down the loop device because it
couldn't find it, and that causes the test to fail. *facepalm*
So, convert all the loop device unmounts to use -d, and direct the
output of all of them to /dev/null.
Dave Chinner [Thu, 26 Jul 2012 08:35:03 +0000 (08:35 +0000)]
xfstests: test 110 sometimes fails to unmount scratch dev
It runs a bunch of rm processes in the background, then immediately
calls _check_scratch_fs without waiting for them to complete, hence
the unmount can fail with a busy filesystem error.
common.attr: get block size of test fs not root fs
stat -f $TEST_DEV shows block size of the fs where $TEST_DEV
resides, usually it is the root fs. This will fail 020 on non-4096 block
size ext2/3/4, also 4096 block size ext2/3/4 on ppc64.
Instead, stat -f $TEST_DIR will show block size of the fs to be tested.
Tested and passed on ext2/3/4, xfs, btrfs with all supported block size.
Cc: Eric Sandeen <esandeen@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
projects / xfstests-dev.git / log